Surface investigations combining the atomic force microscope and scanning electron microscope benefit from the complementarity of both technologies. For the analysis and visualization of a pair of corresponding scans, identifying their spatial alignment is essential. This article presents an automatic registration scheme based on finding correspondences between local image features. The result is then refined by maximizing a similarity measure between the aligned scans. Two feature extraction algorithms have been integrated and tested: Scale-invariant feature transform and speeded-up robust feature. For the experimental evaluation, five types of sample objects showing a variety of different surface structures have been used. The registration scheme succeeds in all application scenarios.